EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: dougduck on June 17, 2024, 08:00:07 pm
-
Hi,
I want to add an LC filter after a DC-DC converter (another board) so as not to change the already designed filter output.
I have been researching some articles about it, and it is always mentioned to add damping to the resonant frequency of the LC circuit.
I'm simulating the behavior, and also with real component values, and it looks pretty good.
BUT... increasing the capacitance value (even with low ESR) seems to eliminate the resonance problem even without damping.
The question is: Is this behavior correct? ???
-
I think you mean "damping".
Yes, that's a good idea.
-
I think you mean "damping".
Yes, that's a good idea.
My question is about the simulation, could using more capacitance eliminate the need for damping? or is this my mistake?
-
The interwinding capacitance real inductor will cause it to be self resonant.
Add a few pf and see the result.
Phase margin? Your circuits have no load. Check the transient response not just the steady state. It's changes in load that will ring the output.
A biggger cap will not help redcuce the Q of the output filter. The aim is to push the Q of the filter down to unity or less.
A second output capacitor with a series resistor will damp down the output filter. This cap should be chosen to be 0.1x the main (low esr) cap.
Tip. Try a putting a resistor in parallel with the inductor. Thats makes for a simple Q killer but at the expense of a bit more ripple.
See attched .
A bigger output cap can cause a surge current and that can lead to converter start up issues.
Damping the input filter of a DC-DC converter matters too.
A DC-DC converter's input behaves as negative resistance. For any given power output, if the input Volts increase, the input current will go down, -V/I.
This adds gain to the input filter and that can lead oscillations.
-
Yes, large enough C, or the appropriate choice of type, can provide damping against L DCR, or its own ESR. Note relying on C ESR introduces a zero to the transfer function, so that it has an LC characteristic at low frequencies, then above the RC or L/R time constant, it has a 1st order (LR) characteristic. We often want some extra (low-ESR) C in parallel to reinforce the 2nd order asymptote at high frequencies. So, it's a common motif, terminating an LC filter with a final C || (R+Cs), where Cs > 3C and R = Zo.
L DCR is impractical for high-current supplies, where an appropriate resistance would dissipate too much power, but it can be acceptable for low-current supplies (local filtering of quiet analog circuitry, say) where the voltage drop is negligible, and "excessively large" C is affordable.
Also for the high-current case, the fact that the filter cap acts in parallel with the overall supply, at frequencies below the LC cutoff, may violate an SMPS maximum-capacitance limits, or draw too much inrush.
Tim